Any idea why bartowski claims DeepSeek-V4-Flash is MXFP4?

https://huggingface.co/bartowski/DeepSeek-V4-Flash-GGUF

This model is in MXFP4 and as such has only been provided in MXFP4 format!

Original model: https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash

But on original's page none of tensors is MXFP4 :

Tensor type BF16 I64 F32 F8_E8M0 F8_E4M3 I8

Am I missing something?